470 AI-acquired users in Month 1, counted the honest way
Before optimizing anything, we built HyNote AI a measurement layer that tells the truth. Server logs showed 92% of raw “AI traffic” was bots. After filtering, what remained was real: 470 AI-acquired users in the first month, roughly 90% of them arriving from ChatGPT.
bot-filtered · Month 1
of AI-acquired users
identified as bots & excluded
Month 1 is a measurement story, on purpose
Most agencies would open with a growth chart. We opened with a filter. HyNote AI came to us as AI answers were becoming a real acquisition channel, and the first job was establishing what was actually true: a locked visibility baseline, a buyer-intent prompt universe, a machine-readable technical layer, and an attribution pipeline that separates humans from bots. Inside that clean measurement, Month 1 counted 470 real AI-acquired users and roughly 135 tracked signups, with about 90% of the AI-acquired users arriving from ChatGPT. No causal growth claims yet — that read comes from the Day 60/90 before-and-after, measured against the baseline we just locked.
Raw analytics were flattering and wrong
For an early-stage SaaS, AI engines are a genuinely new acquisition channel — but the numbers most dashboards report for it are inflated. AI crawlers, scrapers and agent traffic show up in analytics looking like visitors. Optimizing against those numbers means spending real budget on phantom demand.
So before touching content, citations or optimization, the first month answered a harder question: of everything analytics calls “AI traffic,” how much is a human being?
92% of raw AI traffic wasn't human
We cross-checked GA4 against raw server logs, classifying requests by user agent, behavior and origin. The result: roughly 92% of what registered as AI-referred traffic was bot activity — crawlers fetching pages for engines, not buyers arriving from answers. We excluded all of it.
What survived the filter is the number worth reporting: 470 AI-acquired users in Month 1, with ~135 signups tracked, and about 90% of those users arriving from ChatGPT. It is a smaller number than the raw one. It is also the only one that means anything.
Every downstream decision — which prompts to own, which engines to prioritize, what content earns citations — depends on the integrity of this count. An agency reporting unfiltered AI traffic is reporting bots. We count what's real.
Baseline locked, machine layer built, attribution wired
Four workstreams, all in service of clean measurement before optimization.
Three Gates baseline. A locked buyer-intent prompt set establishing where HyNote surfaces today across Retrieval, Recognition and Recommendation — the fixed reference every future lift is measured against.
Buyer-intent prompt universe. Real user language converted into the awareness, consideration and decision prompts HyNote's buyers actually ask AI engines.
Technical layer. Structured data, llms.txt and crawlability work so engines can fetch, understand and quote HyNote's pages cleanly.
Attribution pipeline. GA4 cross-checked against server logs with bot classification — the filter that produced the 470, and the infrastructure every future report runs on.
What we are not claiming
Month 1 makes no causal claim. We are not claiming GEO work drove the 470 users — the baseline was locked during this same window, so a before-and-after doesn't exist yet. We are not claiming visibility lift, and we are not projecting revenue.
The causal read comes next: a Day 60/90 before-and-after against the locked baseline, tracking visibility movement, citation growth and bot-filtered user acquisition on the same prompt set. When we publish that number, you'll know exactly what it was measured against — because the baseline is already on the record.
Request your AI visibility baseline
The first step is a 20–30 minute scoping conversation. We’ll review your category, your competitors, and the prompts your buyers are using, then recommend the right scope.
Book a 30-min call →